The ASTRAL compendium is a collection of software and databases used to aid research into protein structure and evolution. It is complementary to the SCOP: Structural Classification of Proteins database, created by our colleagues at the MRC in Cambridge, UK. SCOP is a comprehensive ordering of domains from all proteins of known structure according to their structural and evolutionary relationships. ASTRAL serves three major purposes: to provide the authoritative sequences and coordinate sets for domains in SCOP, to detect and eliminate errors in the SCOP classification, and to provide additional software and databases to aid in the analysis of the data in SCOP. These additional data include quantitative scores that assign a measure of quality to each crystallographically determined PDB entry in SCOP, representative subsets of PDB chains and SCOP domains, and ASTEROIDS, a system for preliminary classification of domains from newly solved PDB entries within the SCOP framework. ASTRAL development to date has been driven by the needs of our users, who are primarily structural biologists who require SCOP data in easily accessible formats, but lack the resources or expertise to accurately generate the data themselves. Another major group of users are bioinformaticists who use the ASTRAL subsets as a standard representative cross-section of protein structures on which to develop or compare algorithms. A third important group of users are developers of other databases, such as SUPERFAMILY and PRODOM, which are derived from the data in ASTRAL. Our specific aims in this project are to: 1. Produce, develop, and maintain ASTRAL to ensure its long-term utility, in particular: a) Further automate the software used to build ASTRAL. b) Fully document all tools used to build ASTRAL, and make them available as open source. c) Continue to produce new releases of ASTRAL. 2. Extend ASTRAL'S utility to more of the biomedical research community. 3. Develop resources to facilitate extension of ASTRAL by the bioinformatics community.